Irina Topalova - Academia.edu (original) (raw)

Papers by Irina Topalova

Research paper thumbnail of A Model of an Intelligent Automation System for Monitoring of Sensor Signals with a Neural Network Implementation

Annals of DAAAM for ... & proceedings of the ... International DAAAM Symposium .., Dec 31, 2022

The automation systems today combine the capabilities of AI to process large databases in real-ti... more The automation systems today combine the capabilities of AI to process large databases in real-time work, aiming to predict equipment or machine failures. Essential to the reliable and efficient operation of automated systems is the application of AI to monitor their current state. Tracking the status of the sensors in each cycle of machine operation through neural networks would provide an adaptive reflection of faulty or correct behaviour of the automated system. The present study presents a model of an intelligent automation system for monitoring the sensor signals with a neural network implementation. An algorithm for working in two basic modes of a programmable logic controller in this integration is proposed. The neural network is trained with a large number of combinations of sensory signals, corresponding to states of correct behaviour and system faults. Depending on the classification accuracy or currently occurring wrong sensor signals, a retraining method is developed for both modes of operation. The main purpose of the research is to show the effectiveness of the method for classifying the sensor behaviour, in terms of dynamic reaction of the system. The obtained results are discussed and a proposal is made for further development of the research.

Research paper thumbnail of A Conceptual Model for Open U-Learning Platform

Annals of DAAAM for ... & proceedings of the ... International DAAAM Symposium .., Dec 31, 2022

Modern teaching methods applied in a university environment largely determine the quality and eff... more Modern teaching methods applied in a university environment largely determine the quality and effectiveness of the educational process. The choice of a certain method and its application is left to the respective educational institution, which must make the right choice, according to the specifics of its educational programs and goals. In this paper a comparative review of modern e-learning, m-learning and u-learning methods is presented. Their main characteristic parameters are exposed. An open conceptual model for the u-learning platform is proposed. The model is focused on using inside and external, internet-based learning resources and is based on Artificial intelligence to offer the most proper learner-centered learning.

Research paper thumbnail of Dynamic QoS Parameter Adaptation in Routers Using a Multilayer Neural Network

Annals of DAAAM for ... & proceedings of the ... International DAAAM Symposium .., 2019

Practical prevention of network congestion is quality of service (QoS). Connection-oriented proto... more Practical prevention of network congestion is quality of service (QoS). Connection-oriented protocols, such as a TCP protocol, generally look for packet errors, losses, or delays to adjust the transmission speed. Currently, congestion control and avoidance algorithms are based on the idea that packet loss is a suitable indicator of network congestion. The data is transmitted using the Best Effort service model. When a network uses this model, it provides data without guaranteeing reliability and delay. However, these algorithms and the best effort service model are not suitable for applications that respond to delays or packet loss. WRED is implemented on the core routers of a network. The edge routers assign packets with IP priorities when packets enter the network. A neural network is trained to automatically adapt new end users to the quality of service policy, already set by other end-users and accepted by the intermediate routers. The obtained results show that the automated adaptation of the Quality of Service parameters to the already set ones, is possible for the intermediate routers, and the positive cons equences of applying such a method are discussed.

Research paper thumbnail of Automated Marble Plate Classification System Based On Different Neural Network Input Training Sets and PLC Implementation

arXiv (Cornell University), Aug 16, 2012

The process of sorting marble plates according to their surface texture is an important task in t... more The process of sorting marble plates according to their surface texture is an important task in the automated marble plate production. Nowadays some inspection systems in marble industry that automate the classification tasks are too expensive and are compatible only with specific technological equipment in the plant. In this paper a new approach to the design of an Automated Marble Plate Classification System (AMPCS),based on different neural network input training sets is proposed, aiming at high classification accuracy using simple processing and application of only standard devices. It is based on training a classification MLP neural network with three different input training sets: extracted texture histograms, Discrete Cosine and Wavelet Transform over the histograms. The algorithm is implemented in a PLC for real-time operation. The performance of the system is assessed with each one of the input training sets. The experimental test results regarding classification accuracy and quick operation are represented and discussed.

Research paper thumbnail of Classification of ICMP connection time with a multi-layered neural network

Research paper thumbnail of Automated Visual Inspection System in Praline Industry

Annals of DAAAM for ... & proceedings of the ... International DAAAM Symposium .., 2016

There are many industrial applications of sophisticated food packaging systems developed nowadays... more There are many industrial applications of sophisticated food packaging systems developed nowadays. An essential part of these systems is the visual inspection of the package quality. One of the most important tasks in the automated food packaging is the determination of missing elements in the package. The major drawback of these systems is the implementation of too complex techniques, hardware and software methods which raises the cost of the automated system. This paper represents a hardware, software method and algorithm solution for determining missing pralines when automatically loading them into the trays. The main objectives of the research are focused on developing a simple visual method and algorithm using standard hardware components and communication interfaces, easily compatible with different control systems. An image histogram correlation determination is applied on the off-line programing stage. Based on the results, the system is trained with a simple parametrical vector including different histogram parameters to define the number of missing pralines in the tray. The system was tested with different blurred tray images to simulate the movement of the conveyer belt. The results of the obtained high classification accuracy and fast response are represented and discussed. The opportunities for further development of the system are also presented.

Research paper thumbnail of A method for automated classification of steel microstructures based on extraction of informative parameters and neural network implementation

International Conference on Artificial Intelligence, Feb 21, 2009

During the in-service process the structural composition of the steel is changed bringing differe... more During the in-service process the structural composition of the steel is changed bringing different damages which can lead to complete breakdown. It is appropriate to develop methods for automated classification of steel microstructures aiming high recognition accuracy, reliability and lack of any subjective evaluation. The common problem in all of the existing methods for texture classification is the low achieved recognition accuracy rate. That is the reason for searching for new reliable methods giving higher classification accuracy. The goal of the represented research is to propose a new method for automated classification of heat resistant steel structures aiming higher accuracy and computational simplicity in comparison to other existing methods. The proposed method is based on recognition of microscope images for representative steel structures having different aging stage grouped in five classes. In the preprocessing stage the histograms of the images are extracted, stretched and a method for choosing a set of the more informative values of the histogram cover curve is developed. The reduced number of values are given to the input layer neurons of a MLP type neural network. The achieved 100% accuracy and computational simplicity is a good preposition to implement the method for automated calculation of the remaining capacity of the steel avoiding the subjective evaluation factor and implementing it in a real time working systems.

Research paper thumbnail of Modular Adaptive System Based on a Multi-Stage Neural Structure for Recognition of 2D Objects of Discontinuous Production

International Journal of Advanced Robotic Systems, Mar 1, 2005

This is a presentation of a new system for invariant recognition of 2D objects with overlapping c... more This is a presentation of a new system for invariant recognition of 2D objects with overlapping classes, that can not be effectively recognized with the traditional methods. The translation, scale and partial rotation invariant contour object description is transformed in a DCT spectrum space. The obtained frequency spectrums are decomposed into frequency bands in order to feed different BPG neural nets (NNs). The NNs are structured in three stages-filtering and full rotation invariance; partial recognition; general classification. The designed multi-stage BPG Neural Structure shows very good accuracy and flexibility when tested with 2D objects used in the discontinuous production. The reached speed and the opportunuty for an easy restructuring and reprogramming of the system makes it suitable for application in different applied systems for real time work.

Research paper thumbnail of Adaptive approach for filtering the sigma phase in austenitic stainless steel metallographic microstructures

This paper presents an adaptive approach, based on image processing and use of self-organizing ma... more This paper presents an adaptive approach, based on image processing and use of self-organizing maps for filtering, analyzing, and determining the sigma phase percentage in metallographic images of austenitic stainless steel. In order to predict the remaining life of the austenitic stainless steel (12X18H12T), a metallographic analysis of the sigma phase percentage should be made. Following steel microstructure preparation, a series of microscopic digital images are used to measure this parameter. The digital images contain low amount of Gaussian noise and the sigma phase particles must be separated from all non-metal and other small-size or noise inclusions. Implementation of automated measurement leads to more accurate results and minimizes the subjective evaluation factors. A set of morphological features for each blob in a test group of blobs is analyzed using Kohonen self-organizing neural network after applying image filtering and blob detection algorithm. Self-organizing maps are used to filter the blobs. The achieved results are compared with those, obtained from the application of other metallographic methods for the same purpose.

Research paper thumbnail of Increasing the image recognition accuracy in machine vision systems with added noise due to technological issues

ABSTRACT Typical application of machine vision systems in the discrete automated production is qu... more ABSTRACT Typical application of machine vision systems in the discrete automated production is quality control, measurement or classification of moving parts, placed on conveyor belts. Different technical issues (lighting problems, vibrations near camera or conveyor belt, etc.) can lead to noisy images and to wrong classifications or faulty measurements by the vision inspection system. The correlation between motion blur noise (added by technical malfunctions) and the correct measurement by the machine vision system is examined in this paper. First part of the study is to define the influence of motion blur to visual inspection of moving parts with linear velocity of up to 25 m/min. The analyzed vision inspections are size measurement, classification, OCR and code readings. A second study is performed to derive and to propose additional image filtration or vision inspection steps to minimize the wrong measurements according to the inspection type. Of great importance is the added additional amount of processing time. This requires accurate benchmarking of the proposed algorithms within similar laboratory conditions.

Research paper thumbnail of Adaptive Marble Plate Classification System Based on Neural Network and PLC Implementation

Annals of DAAAM for ... & proceedings of the ... International DAAAM Symposium .., 2011

The process of sorting marble plates according to their surface texture is an important task in t... more The process of sorting marble plates according to their surface texture is an important task in the automated marble plate production. Nowadays some inspection systems in marble industry that automates the texture and shade classification tasks are too expensive and are compatible only with specifictechnological equipment in the plant. In this paper a new approach for design of adaptive classification system of marble tiles with similar textures is proposed, aiming at high classification accuracy, applying simple processing and application of only standard devices. It is based on simple image preprocessing, on adaptive training of MLP neural network (MLP NN) with marble histograms and implementation of the algorithm in a Programmable Logic Controller (PLC) for real-time operation. The experimental test results when recognizing marble textures with added motion blur and different illuminations are represented and discussed. The performance of the modeling technique is assessed with different training and test sets. The classification accuracy results are presented and analyzed.

Research paper thumbnail of Optimization of a MLP network through choosing the appropriate input set

Research paper thumbnail of Image and data pre-processing model for real-time communication between dedicated PC and PLC Neural Network application in marble production

A model for image and data pre-processing and communication between a dedicated PC and a PLC with... more A model for image and data pre-processing and communication between a dedicated PC and a PLC with Neural Network (NN) application is proposed in this paper. The proposed model defines guidelines for creating a multithreaded application for receiving real-time data from several digital cameras, parallel image pre-processing based on predefined user algorithms, calculation of input data vector for NN and

Research paper thumbnail of Optimization of a MLP network structure for a real-time PLC application

ABSTRACT The neural networks find many applications today in different kinds of real-time working... more ABSTRACT The neural networks find many applications today in different kinds of real-time working systems. To obtain short execution times and high recognition accuracy in real-time decision-making systems becomes a question of first importance. Therefore, the requirements to the recognition stage in such systems in reference to reduce the reaction time grow up. In the proposed research a new method for optimization of a MLP network structure for a real-time programmable logic controllers (PLC) application is presented. The optimization is accomplished in two steps. First the DCT coefficients are calculated over radial profiles of the objects which form a vector in the frequency parametrical space. This vector describes the corresponding 2D object and is applied as Initial Input Set to the MLP neural network structure. The size of each input for MLP vector is reduced applying modified coefficient of variations (MCV) to evaluate the outlier values. Second the reduced input set is divided and grouped into a number of small MLPs based on analysis of the degree of correlation between the inputs. The trained MLPs are downloaded in a Siemens PLC S7-300 for on-line real-time work in a parallel recognition mode. The proposed optimization is tested for four different 2D objects captured by a CCD matrix camera. The achieved results are represented and analyzed.

Research paper thumbnail of Automated classification of heat resistant steel structures based on neural networks

Gathering enough reliable information for the moment state of metal structures is needed to measu... more Gathering enough reliable information for the moment state of metal structures is needed to measure the rest life time and to assure usage without failures for devices in thermo-electric power plants. This information can be obtained by microstructure analysis of metal specimens by using plastic replicas and structural analysis. During the inservice process (high pressure and temperature) the structural and the phase composition of the steel are changed bringing different damages which can lead to complete breakdown. It is appropriate to develop methods for automated classification of steel structures aiming high recognition accuracy, reliability and lack of any subjective evaluation. The goal of the represented research is to propose a new method for automated classification of heat resistant steel structures aiming higher accuracy in comparison to other existing methods. The proposed method is based on recognition of microscope images for representative steel structures having different aging stage. In the preprocessing stage the histograms of the images are extracted and a set of reduced numbers of the cover curve values are given to the input neurons of a MLP type neural network. The achieved 100% accuracy is a good preposition to implement the method for automated calculation of the remaining capacity of the steel avoiding the subjective evaluation factor and implementing it in a real time working system

Research paper thumbnail of Automated texture classification of marble shades with real-time PLC neural network implementation

ABSTRACT The subjective evaluation of marbles based on their visual appearance could be replaced ... more ABSTRACT The subjective evaluation of marbles based on their visual appearance could be replaced by an automated texture classification system, intending to achieve high classification accuracy and production effectiveness. The existing marble classification methods from a computational point of view are either too complex or very expensive. Nowadays some inspection systems in marble industry that automates the quality-control tasks and shade classification are too expensive and are compatible only with specific technological equipment. In this paper a new approach for classification of marble tiles with similar shades is proposed. It is based on simple image preprocessing, on training a MLP neural network (MLP NN) with marble histograms and implementation of the algorithm in a Programmable Logic Controller (PLC) for real-time execution. A method for training the MLP NN aiming optimization of MLP parameters and topology is proposed. The designed automated system uses only standard PLC modules and communication interfaces. The experimental test results when recognizing marble textures with added motion blur are represented and discussed. The performance of the modeling technique is assessed with different training and test sets. The classification accuracy results are compared to other results obtained by similar approaches.

Research paper thumbnail of Neural Network Structure for Tracking the Climate Temperature Change

Tracking temperature changes in certain geographic regions is a current task in modern research o... more Tracking temperature changes in certain geographic regions is a current task in modern research on Earth's climate changes. One of the global problems in solving this task is related to the large volume of measured data and the search for appropriate methods for effective determination of changes. The purpose of this research is to track climate temperature changes using a machine learning-based automated change detection method. The presented method includes training of a two-level structure of neural networks, with measured temperatures for a ten-year period of time for a certain geographical region. In the testing phase, the neural structure classifies measured temperatures for two three-year periods, before and after the ten-year time period, respectively, for the same geographic region. An algorithm was developed to visualize the studied regions by creating a map with their geographic coordinates. The classification results in the neural structure outputs are presented and ...

Research paper thumbnail of Classification of Two-dimensional Mechanical Parts Using a Convolutional Neural Network

Annals of DAAAM for ... & proceedings of the ... International DAAAM Symposium .., 2022

Image search, object recognition and classification are emerging as key components in modern auto... more Image search, object recognition and classification are emerging as key components in modern automated and autonomous production systems that integrate artificial intelligence. The accuracy in recognizing and classifying these parts, regardless of their geometric transformations, determines to a high degree the accuracy of their manipulation and positioning by the flexible assembly. Based on a comparison of some modern methods for classification of machine parts, the choice of a method with in-depth training of a convolutional neural network is justified. In the presented article a model for classification of machine parts is proposed, which is based on deep training of a convolutional neural network. A model was presented and tested, with a variety of training strategies for the purpose of increased efficiency. The proposed model was based on application of Batch normalization, Gaussian Noise, Weight regularization, Image normalization and Early stopping. High classification accuracy has been achieved for a large training and testing sample. The experiment was conducted for four classes of machine parts having different spatial position and orientation, as well different shapes of the objects belonging to the same class. The parts are grouped in an appropriate training, validation and testing sample. The stability and efficiency of the model under variations of the hyper-parameters of the model have been proven, supported by experimental results.

Research paper thumbnail of Classification of ICMP connection time with a multi-layered neural network

2022 IEEE International Black Sea Conference on Communications and Networking (BlackSeaCom)

Research paper thumbnail of Dynamic QoS Parameter Adaptation in Routers Using a Multilayer Neural Network

DAAAM Proceedings, 2019

Practical prevention of network congestion is quality of service (QoS). Connection-oriented proto... more Practical prevention of network congestion is quality of service (QoS). Connection-oriented protocols, such as a TCP protocol, generally look for packet errors, losses, or delays to adjust the transmission speed. Currently, congestion control and avoidance algorithms are based on the idea that packet loss is a suitable indicator of network congestion. The data is transmitted using the Best Effort service model. When a network uses this model, it provides data without guaranteeing reliability and delay. However, these algorithms and the best effort service model are not suitable for applications that respond to delays or packet loss. WRED is implemented on the core routers of a network. The edge routers assign packets with IP priorities when packets enter the network. A neural network is trained to automatically adapt new end users to the quality of service policy, already set by other end-users and accepted by the intermediate routers. The obtained results show that the automated adaptation of the Quality of Service parameters to the already set ones, is possible for the intermediate routers, and the positive cons equences of applying such a method are discussed.

Research paper thumbnail of A Model of an Intelligent Automation System for Monitoring of Sensor Signals with a Neural Network Implementation

Annals of DAAAM for ... & proceedings of the ... International DAAAM Symposium .., Dec 31, 2022

The automation systems today combine the capabilities of AI to process large databases in real-ti... more The automation systems today combine the capabilities of AI to process large databases in real-time work, aiming to predict equipment or machine failures. Essential to the reliable and efficient operation of automated systems is the application of AI to monitor their current state. Tracking the status of the sensors in each cycle of machine operation through neural networks would provide an adaptive reflection of faulty or correct behaviour of the automated system. The present study presents a model of an intelligent automation system for monitoring the sensor signals with a neural network implementation. An algorithm for working in two basic modes of a programmable logic controller in this integration is proposed. The neural network is trained with a large number of combinations of sensory signals, corresponding to states of correct behaviour and system faults. Depending on the classification accuracy or currently occurring wrong sensor signals, a retraining method is developed for both modes of operation. The main purpose of the research is to show the effectiveness of the method for classifying the sensor behaviour, in terms of dynamic reaction of the system. The obtained results are discussed and a proposal is made for further development of the research.

Research paper thumbnail of A Conceptual Model for Open U-Learning Platform

Annals of DAAAM for ... & proceedings of the ... International DAAAM Symposium .., Dec 31, 2022

Modern teaching methods applied in a university environment largely determine the quality and eff... more Modern teaching methods applied in a university environment largely determine the quality and effectiveness of the educational process. The choice of a certain method and its application is left to the respective educational institution, which must make the right choice, according to the specifics of its educational programs and goals. In this paper a comparative review of modern e-learning, m-learning and u-learning methods is presented. Their main characteristic parameters are exposed. An open conceptual model for the u-learning platform is proposed. The model is focused on using inside and external, internet-based learning resources and is based on Artificial intelligence to offer the most proper learner-centered learning.

Research paper thumbnail of Dynamic QoS Parameter Adaptation in Routers Using a Multilayer Neural Network

Annals of DAAAM for ... & proceedings of the ... International DAAAM Symposium .., 2019

Practical prevention of network congestion is quality of service (QoS). Connection-oriented proto... more Practical prevention of network congestion is quality of service (QoS). Connection-oriented protocols, such as a TCP protocol, generally look for packet errors, losses, or delays to adjust the transmission speed. Currently, congestion control and avoidance algorithms are based on the idea that packet loss is a suitable indicator of network congestion. The data is transmitted using the Best Effort service model. When a network uses this model, it provides data without guaranteeing reliability and delay. However, these algorithms and the best effort service model are not suitable for applications that respond to delays or packet loss. WRED is implemented on the core routers of a network. The edge routers assign packets with IP priorities when packets enter the network. A neural network is trained to automatically adapt new end users to the quality of service policy, already set by other end-users and accepted by the intermediate routers. The obtained results show that the automated adaptation of the Quality of Service parameters to the already set ones, is possible for the intermediate routers, and the positive cons equences of applying such a method are discussed.

Research paper thumbnail of Automated Marble Plate Classification System Based On Different Neural Network Input Training Sets and PLC Implementation

arXiv (Cornell University), Aug 16, 2012

The process of sorting marble plates according to their surface texture is an important task in t... more The process of sorting marble plates according to their surface texture is an important task in the automated marble plate production. Nowadays some inspection systems in marble industry that automate the classification tasks are too expensive and are compatible only with specific technological equipment in the plant. In this paper a new approach to the design of an Automated Marble Plate Classification System (AMPCS),based on different neural network input training sets is proposed, aiming at high classification accuracy using simple processing and application of only standard devices. It is based on training a classification MLP neural network with three different input training sets: extracted texture histograms, Discrete Cosine and Wavelet Transform over the histograms. The algorithm is implemented in a PLC for real-time operation. The performance of the system is assessed with each one of the input training sets. The experimental test results regarding classification accuracy and quick operation are represented and discussed.

Research paper thumbnail of Classification of ICMP connection time with a multi-layered neural network

Research paper thumbnail of Automated Visual Inspection System in Praline Industry

Annals of DAAAM for ... & proceedings of the ... International DAAAM Symposium .., 2016

There are many industrial applications of sophisticated food packaging systems developed nowadays... more There are many industrial applications of sophisticated food packaging systems developed nowadays. An essential part of these systems is the visual inspection of the package quality. One of the most important tasks in the automated food packaging is the determination of missing elements in the package. The major drawback of these systems is the implementation of too complex techniques, hardware and software methods which raises the cost of the automated system. This paper represents a hardware, software method and algorithm solution for determining missing pralines when automatically loading them into the trays. The main objectives of the research are focused on developing a simple visual method and algorithm using standard hardware components and communication interfaces, easily compatible with different control systems. An image histogram correlation determination is applied on the off-line programing stage. Based on the results, the system is trained with a simple parametrical vector including different histogram parameters to define the number of missing pralines in the tray. The system was tested with different blurred tray images to simulate the movement of the conveyer belt. The results of the obtained high classification accuracy and fast response are represented and discussed. The opportunities for further development of the system are also presented.

Research paper thumbnail of A method for automated classification of steel microstructures based on extraction of informative parameters and neural network implementation

International Conference on Artificial Intelligence, Feb 21, 2009

During the in-service process the structural composition of the steel is changed bringing differe... more During the in-service process the structural composition of the steel is changed bringing different damages which can lead to complete breakdown. It is appropriate to develop methods for automated classification of steel microstructures aiming high recognition accuracy, reliability and lack of any subjective evaluation. The common problem in all of the existing methods for texture classification is the low achieved recognition accuracy rate. That is the reason for searching for new reliable methods giving higher classification accuracy. The goal of the represented research is to propose a new method for automated classification of heat resistant steel structures aiming higher accuracy and computational simplicity in comparison to other existing methods. The proposed method is based on recognition of microscope images for representative steel structures having different aging stage grouped in five classes. In the preprocessing stage the histograms of the images are extracted, stretched and a method for choosing a set of the more informative values of the histogram cover curve is developed. The reduced number of values are given to the input layer neurons of a MLP type neural network. The achieved 100% accuracy and computational simplicity is a good preposition to implement the method for automated calculation of the remaining capacity of the steel avoiding the subjective evaluation factor and implementing it in a real time working systems.

Research paper thumbnail of Modular Adaptive System Based on a Multi-Stage Neural Structure for Recognition of 2D Objects of Discontinuous Production

International Journal of Advanced Robotic Systems, Mar 1, 2005

This is a presentation of a new system for invariant recognition of 2D objects with overlapping c... more This is a presentation of a new system for invariant recognition of 2D objects with overlapping classes, that can not be effectively recognized with the traditional methods. The translation, scale and partial rotation invariant contour object description is transformed in a DCT spectrum space. The obtained frequency spectrums are decomposed into frequency bands in order to feed different BPG neural nets (NNs). The NNs are structured in three stages-filtering and full rotation invariance; partial recognition; general classification. The designed multi-stage BPG Neural Structure shows very good accuracy and flexibility when tested with 2D objects used in the discontinuous production. The reached speed and the opportunuty for an easy restructuring and reprogramming of the system makes it suitable for application in different applied systems for real time work.

Research paper thumbnail of Adaptive approach for filtering the sigma phase in austenitic stainless steel metallographic microstructures

This paper presents an adaptive approach, based on image processing and use of self-organizing ma... more This paper presents an adaptive approach, based on image processing and use of self-organizing maps for filtering, analyzing, and determining the sigma phase percentage in metallographic images of austenitic stainless steel. In order to predict the remaining life of the austenitic stainless steel (12X18H12T), a metallographic analysis of the sigma phase percentage should be made. Following steel microstructure preparation, a series of microscopic digital images are used to measure this parameter. The digital images contain low amount of Gaussian noise and the sigma phase particles must be separated from all non-metal and other small-size or noise inclusions. Implementation of automated measurement leads to more accurate results and minimizes the subjective evaluation factors. A set of morphological features for each blob in a test group of blobs is analyzed using Kohonen self-organizing neural network after applying image filtering and blob detection algorithm. Self-organizing maps are used to filter the blobs. The achieved results are compared with those, obtained from the application of other metallographic methods for the same purpose.

Research paper thumbnail of Increasing the image recognition accuracy in machine vision systems with added noise due to technological issues

ABSTRACT Typical application of machine vision systems in the discrete automated production is qu... more ABSTRACT Typical application of machine vision systems in the discrete automated production is quality control, measurement or classification of moving parts, placed on conveyor belts. Different technical issues (lighting problems, vibrations near camera or conveyor belt, etc.) can lead to noisy images and to wrong classifications or faulty measurements by the vision inspection system. The correlation between motion blur noise (added by technical malfunctions) and the correct measurement by the machine vision system is examined in this paper. First part of the study is to define the influence of motion blur to visual inspection of moving parts with linear velocity of up to 25 m/min. The analyzed vision inspections are size measurement, classification, OCR and code readings. A second study is performed to derive and to propose additional image filtration or vision inspection steps to minimize the wrong measurements according to the inspection type. Of great importance is the added additional amount of processing time. This requires accurate benchmarking of the proposed algorithms within similar laboratory conditions.

Research paper thumbnail of Adaptive Marble Plate Classification System Based on Neural Network and PLC Implementation

Annals of DAAAM for ... & proceedings of the ... International DAAAM Symposium .., 2011

The process of sorting marble plates according to their surface texture is an important task in t... more The process of sorting marble plates according to their surface texture is an important task in the automated marble plate production. Nowadays some inspection systems in marble industry that automates the texture and shade classification tasks are too expensive and are compatible only with specifictechnological equipment in the plant. In this paper a new approach for design of adaptive classification system of marble tiles with similar textures is proposed, aiming at high classification accuracy, applying simple processing and application of only standard devices. It is based on simple image preprocessing, on adaptive training of MLP neural network (MLP NN) with marble histograms and implementation of the algorithm in a Programmable Logic Controller (PLC) for real-time operation. The experimental test results when recognizing marble textures with added motion blur and different illuminations are represented and discussed. The performance of the modeling technique is assessed with different training and test sets. The classification accuracy results are presented and analyzed.

Research paper thumbnail of Optimization of a MLP network through choosing the appropriate input set

Research paper thumbnail of Image and data pre-processing model for real-time communication between dedicated PC and PLC Neural Network application in marble production

A model for image and data pre-processing and communication between a dedicated PC and a PLC with... more A model for image and data pre-processing and communication between a dedicated PC and a PLC with Neural Network (NN) application is proposed in this paper. The proposed model defines guidelines for creating a multithreaded application for receiving real-time data from several digital cameras, parallel image pre-processing based on predefined user algorithms, calculation of input data vector for NN and

Research paper thumbnail of Optimization of a MLP network structure for a real-time PLC application

ABSTRACT The neural networks find many applications today in different kinds of real-time working... more ABSTRACT The neural networks find many applications today in different kinds of real-time working systems. To obtain short execution times and high recognition accuracy in real-time decision-making systems becomes a question of first importance. Therefore, the requirements to the recognition stage in such systems in reference to reduce the reaction time grow up. In the proposed research a new method for optimization of a MLP network structure for a real-time programmable logic controllers (PLC) application is presented. The optimization is accomplished in two steps. First the DCT coefficients are calculated over radial profiles of the objects which form a vector in the frequency parametrical space. This vector describes the corresponding 2D object and is applied as Initial Input Set to the MLP neural network structure. The size of each input for MLP vector is reduced applying modified coefficient of variations (MCV) to evaluate the outlier values. Second the reduced input set is divided and grouped into a number of small MLPs based on analysis of the degree of correlation between the inputs. The trained MLPs are downloaded in a Siemens PLC S7-300 for on-line real-time work in a parallel recognition mode. The proposed optimization is tested for four different 2D objects captured by a CCD matrix camera. The achieved results are represented and analyzed.

Research paper thumbnail of Automated classification of heat resistant steel structures based on neural networks

Gathering enough reliable information for the moment state of metal structures is needed to measu... more Gathering enough reliable information for the moment state of metal structures is needed to measure the rest life time and to assure usage without failures for devices in thermo-electric power plants. This information can be obtained by microstructure analysis of metal specimens by using plastic replicas and structural analysis. During the inservice process (high pressure and temperature) the structural and the phase composition of the steel are changed bringing different damages which can lead to complete breakdown. It is appropriate to develop methods for automated classification of steel structures aiming high recognition accuracy, reliability and lack of any subjective evaluation. The goal of the represented research is to propose a new method for automated classification of heat resistant steel structures aiming higher accuracy in comparison to other existing methods. The proposed method is based on recognition of microscope images for representative steel structures having different aging stage. In the preprocessing stage the histograms of the images are extracted and a set of reduced numbers of the cover curve values are given to the input neurons of a MLP type neural network. The achieved 100% accuracy is a good preposition to implement the method for automated calculation of the remaining capacity of the steel avoiding the subjective evaluation factor and implementing it in a real time working system

Research paper thumbnail of Automated texture classification of marble shades with real-time PLC neural network implementation

ABSTRACT The subjective evaluation of marbles based on their visual appearance could be replaced ... more ABSTRACT The subjective evaluation of marbles based on their visual appearance could be replaced by an automated texture classification system, intending to achieve high classification accuracy and production effectiveness. The existing marble classification methods from a computational point of view are either too complex or very expensive. Nowadays some inspection systems in marble industry that automates the quality-control tasks and shade classification are too expensive and are compatible only with specific technological equipment. In this paper a new approach for classification of marble tiles with similar shades is proposed. It is based on simple image preprocessing, on training a MLP neural network (MLP NN) with marble histograms and implementation of the algorithm in a Programmable Logic Controller (PLC) for real-time execution. A method for training the MLP NN aiming optimization of MLP parameters and topology is proposed. The designed automated system uses only standard PLC modules and communication interfaces. The experimental test results when recognizing marble textures with added motion blur are represented and discussed. The performance of the modeling technique is assessed with different training and test sets. The classification accuracy results are compared to other results obtained by similar approaches.

Research paper thumbnail of Neural Network Structure for Tracking the Climate Temperature Change

Tracking temperature changes in certain geographic regions is a current task in modern research o... more Tracking temperature changes in certain geographic regions is a current task in modern research on Earth's climate changes. One of the global problems in solving this task is related to the large volume of measured data and the search for appropriate methods for effective determination of changes. The purpose of this research is to track climate temperature changes using a machine learning-based automated change detection method. The presented method includes training of a two-level structure of neural networks, with measured temperatures for a ten-year period of time for a certain geographical region. In the testing phase, the neural structure classifies measured temperatures for two three-year periods, before and after the ten-year time period, respectively, for the same geographic region. An algorithm was developed to visualize the studied regions by creating a map with their geographic coordinates. The classification results in the neural structure outputs are presented and ...

Research paper thumbnail of Classification of Two-dimensional Mechanical Parts Using a Convolutional Neural Network

Annals of DAAAM for ... & proceedings of the ... International DAAAM Symposium .., 2022

Image search, object recognition and classification are emerging as key components in modern auto... more Image search, object recognition and classification are emerging as key components in modern automated and autonomous production systems that integrate artificial intelligence. The accuracy in recognizing and classifying these parts, regardless of their geometric transformations, determines to a high degree the accuracy of their manipulation and positioning by the flexible assembly. Based on a comparison of some modern methods for classification of machine parts, the choice of a method with in-depth training of a convolutional neural network is justified. In the presented article a model for classification of machine parts is proposed, which is based on deep training of a convolutional neural network. A model was presented and tested, with a variety of training strategies for the purpose of increased efficiency. The proposed model was based on application of Batch normalization, Gaussian Noise, Weight regularization, Image normalization and Early stopping. High classification accuracy has been achieved for a large training and testing sample. The experiment was conducted for four classes of machine parts having different spatial position and orientation, as well different shapes of the objects belonging to the same class. The parts are grouped in an appropriate training, validation and testing sample. The stability and efficiency of the model under variations of the hyper-parameters of the model have been proven, supported by experimental results.

Research paper thumbnail of Classification of ICMP connection time with a multi-layered neural network

2022 IEEE International Black Sea Conference on Communications and Networking (BlackSeaCom)

Research paper thumbnail of Dynamic QoS Parameter Adaptation in Routers Using a Multilayer Neural Network

DAAAM Proceedings, 2019

Practical prevention of network congestion is quality of service (QoS). Connection-oriented proto... more Practical prevention of network congestion is quality of service (QoS). Connection-oriented protocols, such as a TCP protocol, generally look for packet errors, losses, or delays to adjust the transmission speed. Currently, congestion control and avoidance algorithms are based on the idea that packet loss is a suitable indicator of network congestion. The data is transmitted using the Best Effort service model. When a network uses this model, it provides data without guaranteeing reliability and delay. However, these algorithms and the best effort service model are not suitable for applications that respond to delays or packet loss. WRED is implemented on the core routers of a network. The edge routers assign packets with IP priorities when packets enter the network. A neural network is trained to automatically adapt new end users to the quality of service policy, already set by other end-users and accepted by the intermediate routers. The obtained results show that the automated adaptation of the Quality of Service parameters to the already set ones, is possible for the intermediate routers, and the positive cons equences of applying such a method are discussed.